[SCM] Socket Wrapper Repository - branch master updated
The branch, master has been updated
via 13a6aca swrap: fix copy on write leak of ~38M for every fork.
via 92a4fce swrap: abort on mutex errors
via affaf42 swrap: fallback to libc_getpeername() when we get an empty
sun_path from accept()
via e72898a swrap: make swrap_accept() more resilient against races
related to already disconnected sockets
via c3f7465 swrap: add better logging to convert_un_in()
via fa7a9b7 tests/echo_srv: make the main server logic resilient to
ECONNABORTED from accept()
from cd51d80 Bump version to 1.3.0
https://git.samba.org/?p=socket_wrapper.git;a=shortlog;h=master
- Log -
commit 13a6aca342120383776251247c72170c142bf017
Author: Stefan Metzmacher
Date: Thu Feb 4 05:02:32 2021 +0100
swrap: fix copy on write leak of ~38M for every fork.
commit 0f8e90dd7e59c473be615dee08d445dca98fdab9
(src/socket_wrapper.c: fix mutex fork handling)
let us touch the whole sockets array on every fork,
because each element in the array has it's own mutex.
max_sockets=65535 * sizeof(struct socket_info_container)=592 = 38796720
This was designed for the use of robust shared mutexes
when moving the sockets array into a shared memory file.
Until we really move to shared memory, we can use a single
global mutex in order to avoid the copy on write leaking.
Signed-off-by: Stefan Metzmacher
Reviewed-by: Andreas Schneider
commit 92a4fce9df55e4c1a3b75325baa0f4b0988eeb6c
Author: Stefan Metzmacher
Date: Fri Feb 5 11:50:17 2021 +0100
swrap: abort on mutex errors
There's no way to continue in a reliable way...
Signed-off-by: Stefan Metzmacher
Reviewed-by: Andreas Schneider
commit affaf4248c0cc2056081f56199109e4a94348b82
Author: Stefan Metzmacher
Date: Fri Feb 5 13:13:44 2021 +0100
swrap: fallback to libc_getpeername() when we get an empty sun_path from
accept()
This hopefully hides the strange behaviour of FreeBSD (at least 12.1)
for already disconnected AF_UNIX sockets.
The race is triggered when the following detects the usage of 'getpeername':
truss -o ./truss.out -f -H -a -e -D -s 160 ctest -V -R
test_thread_echo_tcp_connect;
grep getpeername truss.out
In a simplified log the following is happening:
ECHO_SRV(parent): socket(PF_LOCAL,SOCK_STREAM,0) = 4 (0x4)
ECHO_SRV(parent): unlink("/tmp/w_E37bkf/T0A0007") ERR#2 'No such file or
directory'
ECHO_SRV(parent): bind(4,{ AF_UNIX "/tmp/w_E37bkf/T0A0007" },106) = 0 (0x0)
ECHO_SRV(parent): listen(4,16) = 0 (0x0)
...
ECHO_SRV(parent): write(2,"SWRAP_ERROR[echo_srv (9792)] - swrap_accept:
before accept(sa_socklen=106)\n",75) = 75 (0x4b)
ECHO_SRV(parent): accept4(0x4,0x7ffde158,0x7ffde150,0x0) = 5 (0x5)
ECHO_SRV(parent): write(2,"SWRAP_ERROR[echo_srv (9792)] - swrap_accept:
after accept(sa_socklen=106, family=1)\n",84) = 84 (0x54)
ECHO_SRV(parent): getsockname(5,{ AF_UNIX "/tmp/w_E37bkf/T0A0007"
},0x7ffde0c0) = 0 (0x0)
ECHO_SRV(parent): swrap_accept() returned a valid connection and a per
connection child (pid=9793) handles it
TEST_THREAD: socket(PF_LOCAL,SOCK_STREAM,0) = 7 (0x7)
TEST_THREAD: bind(7,{ AF_UNIX "/tmp/w_E37bkf/T014D4F" },106) = 0 (0x0)
TEST_THREAD: connect(7,{ AF_UNIX "/tmp/w_E37bkf/T0A0007" },106) = 0
(0x0)
TEST_THREAD: close(7) = 0 (0x0)
ECHO_SRV(parent): wait4(-1,0x0,0x0,0x0) = 9793 (0x2641)
ECHO_SRV(parent): close(5) = 0 (0x0)
ECHO_SRV(parent): write(2,"SWRAP_ERROR[echo_srv (9792)] - swrap_accept:
before accept(sa_socklen=106)\n",75) = 75 (0x4b)
ECHO_SRV(parent): accept4(0x4,0x7ffde158,0x7ffde150,0x0) = 5 (0x5)
TEST_THREAD: unlink("/tmp/w_E37bkf/T014D4F") = 0 (0x0)
ECHO_SRV(parent): write(2,"SWRAP_ERROR[echo_srv (9792)] - swrap_accept:
after accept(sa_socklen=16, family=1)\n",83) = 83 (0x53)
ECHO_SRV(parent): getpeername(5,0x7ffde158,0x7ffde150) ERR#57
'Socket is not connected'
ECHO_SRV(parent): getsockname(5,{ AF_UNIX "/tmp/w_E37bkf/T0A0007"
},0x7ffde0c0) = 0 (0x0)
ECHO_SRV(parent): getpeername(5,0x7ffde158,0x7ffde150) ERR#57
'Socket is not connected'
Signed-off-by: Stefan Metzmacher
Reviewed-by: Andreas Schneider
commit e72898ad92a52a595d4733210483e9689cb5d390
Author: Stefan Metzmacher
Date: Fri Feb 5 12:13:12 2021 +0100
swrap: make swrap_accept() more resilient against races related to already
disconnected sockets
Callers of accept() expect to get ECONNABORTED instead of a disconnected
socket.
Even on Linux we have a potential race calling libc_getsockname()
after accept(), so we map ENOTCONN to ECONNABORTED.
We should do all syscalls in order to have peer
[SCM] Samba Shared Repository - branch v4-12-test updated
The branch, v4-12-test has been updated via 40e503732dc lib:util: Avoid free'ing our own pointer via 65b3648c553 lib:util: Add cache oversize test for memcache via a658dd19627 lib:util: Add basic memcache unit test via 6e6aa90b87b s3: libsmb: cli_state_save_tcon(). Don't deepcopy tcon struct when temporarily swapping out a connection on a cli_state. via bab7f2ae28e s3: torture: Change the SMB1-only UID-REGRESSION-TEST to do an explicit copy of the tcon struct in use. via a19f94c644d s3: smbtorture3: Ensure run_tcon_test() always replaces any saved tcon and shuts down correctly even in error paths. via 80d2c3e4725 s3: smbtorture3: Ensure we *always* replace the saved saved_tcon even in an error condition. via a00ff434515 s3: tests: Add regression test for bug 13992. from 8c846741a45 s3/auth: implement "winbind:ignore domains" https://git.samba.org/?p=samba.git;a=shortlog;h=v4-12-test - Log - commit 40e503732dc0129b8e5f7bc84ca027514d2ef02e Author: Andreas Schneider Date: Tue Feb 2 18:10:38 2021 +0100 lib:util: Avoid free'ing our own pointer BUG: https://bugzilla.samba.org/show_bug.cgi?id=14625 Signed-off-by: Andreas Schneider Reviewed-by: Ralph Boehme (cherry picked from commit 0bdbe50fac680be3fe21043246b8c75005611351) Autobuild-User(v4-12-test): Karolin Seeger Autobuild-Date(v4-12-test): Fri Feb 5 12:14:09 UTC 2021 on sn-devel-184 commit 65b3648c55332d6186c6a03c0ef56e683a7afa46 Author: Andreas Schneider Date: Wed Feb 3 10:37:12 2021 +0100 lib:util: Add cache oversize test for memcache BUG: https://bugzilla.samba.org/show_bug.cgi?id=14625 Signed-off-by: Andreas Schneider Reviewed-by: Ralph Boehme (cherry picked from commit 00543ab3b29e3fbfe8314e51919629803e14ede6) commit a658dd19627ebe43d4e4fc494ff0d60507554a84 Author: Andreas Schneider Date: Wed Feb 3 10:30:08 2021 +0100 lib:util: Add basic memcache unit test BUG: https://bugzilla.samba.org/show_bug.cgi?id=14625 Signed-off-by: Andreas Schneider Reviewed-by: Ralph Boehme (cherry picked from commit bebbf621d6052f797c5cf19a2a9bbc13e699d3f0) commit 6e6aa90b87b5ddaa6b92160eb467090705e25ff6 Author: Jeremy Allison Date: Thu Jan 28 11:08:48 2021 -0800 s3: libsmb: cli_state_save_tcon(). Don't deepcopy tcon struct when temporarily swapping out a connection on a cli_state. This used to make a deep copy of either cli->smb2.tcon or cli->smb1.tcon, but this leaves the original tcon pointer in place which will then get TALLOC_FREE()'d when the new tree connection is made on this cli_state. As there may be pipes open on the old tree connection with talloc'ed state allocated using the original tcon pointer as a talloc parent we can't deep copy and then free this pointer as that will fire the destructors on the pipe memory and mark them as not connected. This call is used to temporarily swap out a tcon pointer (whilst keeping existing pipes open) to allow a new tcon on the same cli_state and all users correctly call cli_state_restore_tcon() once they are finished with the new tree connection. Just return the existing pointer and set the old value to NULL. We know we MUST be calling cli_state_restore_tcon() below to restore the original tcon tree connection pointer before closing the session. Remove the knownfail.d entry. BUG: https://bugzilla.samba.org/show_bug.cgi?id=13992 Signed-off-by: Jeremy Allison Reviewed-by: Andreas Schneider Autobuild-User(master): Jeremy Allison Autobuild-Date(master): Tue Feb 2 21:05:25 UTC 2021 on sn-devel-184 (cherry picked from commit 4f80f5f9046b64a9e5e0503b1cb54f1492c4faec) commit bab7f2ae28e5bd0a44e121c9b7e9d868271fac70 Author: Jeremy Allison Date: Thu Jan 28 17:35:55 2021 -0800 s3: torture: Change the SMB1-only UID-REGRESSION-TEST to do an explicit copy of the tcon struct in use. For this test only, explicitly copy the SMB1 tcon struct, don't use cli_state_save_tcon()//cli_state_restore_tcon() as these calls will soon change to just manipulate the pointer to avoid TALLOC_FREE() on the tcon struct which calls destructors on child pipe data. In SMB1 this test calls cli_tdis() twice with an invalid vuid and expects the SMB1 tcon struct to be preserved across the calls. SMB1 cli_tdis() frees cli->smb1.tcon so we must put back a deep copy into cli->smb1.tcon to be able to safely call cli_tdis() again. This is a test-only hack. Real client code uses cli_state_save_tcon()/cli_state_restore_tcon() if it needs to temporarily swap out the active tcon on a client connection. BUG: https://bugzilla.samba.org/show_bug.cgi?id=13992
